Students' performance dataset for using machine learning technique in physics education research.

Area of Science:

• Physics Education Research
• Educational Data Mining
• Machine Learning Applications

Background:

• Challenges exist in advancing machine learning and data mining in physics education research due to a lack of specific datasets.
• Existing methods for predicting student performance in physics education research may not be optimal.

Purpose of the Study:

• To introduce the Students' Performance Dataset in Physics Education Research (SPHERE) dataset.
• To demonstrate the utility of the SPHERE dataset for training machine learning models.
• To compare the predictive performance of machine learning models trained on SPHERE with traditional teacher-based assessments.

Main Methods:

• Collected physics performance data from students across three domains: conceptual understanding, scientific ability, and learning attitude.
• Utilized research-based assessments (RBAs) aligned with the curriculum for eleventh-grade physics.
• Applied machine learning techniques to the SPHERE dataset for performance prediction.

Main Results:

• The SPHERE dataset provides valuable data for physics education research.
• Machine learning models trained on SPHERE demonstrated superior predictive performance.
• SPHERE-based predictions outperformed those made by physics teachers.

Conclusions:

• The SPHERE dataset is a significant resource for advancing machine learning applications in physics education research.
• Machine learning models utilizing the SPHERE dataset offer a more accurate method for predicting student physics performance.
• This work highlights the potential of data-driven approaches in improving physics education research and practice.